The Next Frontier for AI: Powering Intelligence at the Edge - Six Five On The Road

How are organizations and technology leaders preparing for AI’s rapid movement from centralized data centers to the intelligent edge?

‍

From Cisco’s Partner Summit 2025, host Daniel Newman is joined by Cisco's President & Chief Product Officer, Jeetu Patel, for a conversation on how Cisco’s Unified Edge is enabling AI to operate securely and efficiently beyond the confines of the data center. The discussion focuses on the explosion of AI infrastructure innovation, the challenges organizations face when moving from pilot projects to full-scale deployment, and the implications of bringing AI-powered intelligence to physical environments and distributed locations.

‍

Key Takeaways Include:

‍

🔹Explosion in AI infrastructure: New advances are surpassing the innovation seen in the last decade and redefining how organizations approach investment planning for AI.

‍

🔹Bridging the pilot-to-production gap: Simplifying AI deployment, management, and security is crucial to enabling organizations to scale from testing to tangible results.

‍

🔹Edge intelligence in practice: The shift to distributed and Edge AI is driven by needs for improved performance, reduced latency, and greater cost efficiency, impacting a broad range of industries from retail to manufacturing.

‍

🔹Real-world impact and differentiation: Cisco’s approach to Edge AI emphasizes secure, scalable, and seamless integration compared to traditional hyperscaler or startup methods, changing how connected experiences are delivered.

‍

🔹Navigating new dynamics: As AI becomes pervasive across various locations, considerations around experience, privacy, control, and the pivotal role of the network are shaping the next phase of AI adoption.

‍

Learn more at Cisco.com

‍

Watch the full video at sixfivemedia.com, and be sure to subscribe to our YouTube channel so you never miss an episode.

‍

Or listen to the audio here:

‍

Disclaimer: Six Five On The Road is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded, and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors, and we ask that you do not treat us as such.

‍

Daniel Newman: Hey everyone. The Six Five is On the Road. We're here in San Diego and I'm really excited for this conversation. We're having a return guest, a regular here on the Six five, someone I really enjoy talking to quite a bit. I've got Jeetu.

‍

Jeetu Patel: How are you, my friend?

‍

Daniel Newman: It's good. It's always good to sit down with you. You know, one of the things that's been on my mind a lot lately is, you know, we hear so much about AI. G2 all. It's so focused on the consumer. It's so focused on chatbots. Right. And like one of the worst arguments in the market is that AI is a bubble because of how much revenue chatbots are creating right now. You know, you and I have been around this for decades and AI has been seeping its way into enterprises for years. We had the big data revolution, the machine learning revolutions of the past years. We've been talking for two decades about building infrastructure, network storage capacity, security for all your data. Because we knew in the future you'd be able to do this. Incredible things. I'd love to get your read on sort of where we're at with enterprise AI because my feeling is we are so early and when people are looking at the investments that are being made and the spend being made for these data centers, it's not about this consumer chatbot era. It's about the enterprises. When they start adopting this stuff, that's where the numbers start to get really, really exciting.

‍

Jeetu Patel: I think it's on both, frankly. I think the consumer side is also.

‍

Daniel Newman: I should say, just the consumer, not just the consumer side.

‍

Jeetu Patel: I'd completely agree with you. If I were to say what seems counterintuitive right now, but is actually completely factually correct, is that we are underestimating the demand signal and that the data center build outs and the trillions of dollars with gigawatts of capacity is still actually under what is going to be needed to go out and fulfill the demands of AI as we move forward. And that is counterintuitive for most people as they think about this. I think you'll have. I was just in the Middle East and we have a massive shortage on three fronts. Right. We've got a shortage of infrastructure, a shortage of trust and security, where there's a trust deficit and then there's a data shortage. There's a data gap of some sort.

‍

Daniel Newman: And energy, I would argue energy on.

‍

Jeetu Patel: The infrastructure side, the most scarce resource.

‍

Daniel Newman: Yeah.

‍

Jeetu Patel: And the big constraint is power.

‍

Daniel Newman: Yeah, there you go.

‍

Jeetu Patel: The asset is The GPU and the force multiplier is the network. You should think about it that way. But I feel like we are underestimating the demand. And the reason that even on the consumer side you're not seeing more usage is not because there's not more demand, it's because they don't have enough computers to be able to turn on everyone. So, like, if you think about Sora 2, Sora 2 could be using a lot more compute if they just have the availability. They just don't have the availability.

‍

Daniel Newman: They have a lot more paying customers, right?

‍

Jeetu Patel: They have way more paying customers than what they do right now. So I feel like there's going to be an exponentiality in the curve on revenue that people are underestimating. And then when you go to the enterprise, right now we're in the experimentation phase for the most part, a lot of experiments running. You're now starting to see certain kinds of areas of the enterprise, like coding and like, you know, kind of contact centers and customer success. Those kinds of areas are starting to pick up. But in the next two years, three years, every pharmaceutical company, every financial services company, every healthcare company, you'll see a very, very different level of usage pattern for AI.

‍

Daniel Newman: It's very interesting. I've been trying to, to explain, and you and I spend a little time on social, we spend time in media as well. And like, I've got these two schools, I've got these apps, I call them the Bubble Bears. The ones that are like, it's just all a bubble. It's all this big circular thing. And it's just four or five companies all buying from each other. And then there's obviously the others that like you and I that tend to be like, no, this is pretty exponential. I love the point you made about, you know, if you listen to Sam Altman, you listen to Elon Musk, they may not get along in real life, but they both will say the same thing. They don't have enough compute. And then if you look. What is it? I woke up yesterday, literally from the time I woke up in Austin, flew to Washington D.C. i woke up and there was over $70 billion of new AI deals that were done in one day. There was a Amazon aws deal with OpenAI. There was a deal in West Texas with a company called Cypher. And AWS, there was a Microsoft deal with Iron Limited. And I think Microsoft also got some dollars to go to the Middle east now and start deploying, which is going to be a huge opportunity.

‍

Jeetu Patel: We're doing the same one day, by.

‍

Daniel Newman: The way though, Jeetu, one day, $70 billion. And what it really comes down to is that we have to build these things first. And so what I keep saying, I'd love to get your take on this, is that what you're thinking and what other CEOs, presidents of these companies, whether it's Zuck, whether it's Satya, whether it's Andy Jassy, whether it's Sundar, what they're all basically saying is that it's an existential risk to their business to under build. I think Zuck says I would rather over invest than underinvest because you absolutely cannot afford to get this wrong. So do you agree that basically, with all I know there's a lot of fear of capex, I know there's a lot of are we overspending? But is the risk bigger for these companies that are trying to make sure they maintain, including Cisco, their place to potentially invest is a safer play than under investing now?

‍

Jeetu Patel: I think it's a normal human tendency. We always assess the risk of doing something. Yeah, we never assess the risk of not doing something. And the risk of not doing something sometimes can be far greater than the risk of doing something and being wrong. And so what's the risk of being wrong? If you've overbuilt capacity, it takes you a while to kind of grow into it. What's the risk of underinvesting? You're literally not in the conversation and you missed the mark, you missed the wave and someone else took over and took your share and you are now no longer relevant in the next wave and your profit pool shrank quite materially as a result of that. So I agree with you. The thing that I think we have to keep in mind is that in the short term, most of these waves tend to overestimate the impact. In the long term, people grossly underestimate it. And so in the next six months, agentic AI might not transform every business process on the planet, but in the next five years or the next seven years, it's very hard to imagine not having every process in the business get re-engineered and rethought through. And you have to plan for that. One of the things I tell my team is because we are doing the coding piece and initially you get a lot of resistance from people saying, oh, the technology's not ready. And what I've told them is don't think about the technology not being ready today, think about what this will do in three months from now. Not three years. Three months from now, because in three months it will be materially different than what it is today. And at that point you have to get yourself in a mental model that in three months you will be the leader when the technology reaches that point. And that's the mental model that we have to actually operate with.

‍

Daniel Newman: It's interesting, you've probably heard this in leadership a number of times, but people tend to overestimate what they can accomplish in a year and grossly underestimate what.

‍

Jeetu Patel: They can do it at five.

‍

Daniel Newman: Yeah, it's the same thing though, with this technology boom. But one of the other things is just the speed and proliferation. I think Greg Brockman said this. He was looking at the proliferation of just OpenAI and ChatGPT, but I think he said it was seven times faster than the Internet. So one of the other things though is just the build to speed is that that gap, Jeetu, like the gap from the time that you're going to deploy the infrastructure and start with say an agentic project, start with a customer success project, start with a coding project to when you know however good it is today is the absolute worst it's ever going to be. I mean, you have to be seeing that, like even things that six months ago you looked at your team and said, yeah, we can use code to support 20%. What's that like? What's that say? How fast is that proliferating as you know, within these different things you're trying to take off?

‍

Jeetu Patel: So I think projecting exponentiality is really hard for the collective, you know, for sure kind of corpus of humans. So what we have to do is we have to keep reminding ourselves that the pace is only going to accelerate, not decelerate. And just because something's not ready right now, if you can see a germ of a seed that tells you that it's going to be ready in a few months, you will be surprised because it'll be ready faster than you think. And you have to get prepared at that point in time. So one of the things that I've always felt is the direct benefactors of AI to date are the ones that didn't sit on the sidelines, but actually jumped in and understood, experimented and owned their own kind of destiny moving forward. And so I feel it's like a really bad strategy to just sit, observe and say, when it gets ready, I'll actually be. I'll take it up at that point in time. I think that's a really bad strategy that almost guarantees failure.

Daniel Newman: Absolutely. My read on it is just when people like to use the argument that AI can't do this capability yet. And I think the word yet is very important because of how fast things are coming to fruition. And I think that's one of my biggest points. Now, in terms of Cisco, where do you sort of sit right now in this? I mean, you guys have a huge footprint in security. Obviously. You made a great point about networking being one of the great enablers, right? You can have all the compute in the world, you have to connect all the, make all the data move. Security. I mean, where are you sitting? Where are you seeing the biggest momentum? Where are you getting the biggest pickup? I know you got that billion plus order book. It's growing really quickly.

‍

Jeetu Patel: We said it was a billion. It actually ended up being north of 2. And we're in the quiet period right now. Earnings come out next quarter, so we'll tell you what happens then. But it's in general, I'd say that, you know, at Cisco, we're lucky because we're the core of the three big constraints that you have in AI. We're the core of the infrastructure constraint, because if you have a warehouse full of GPUs that don't get networked, they're no good to you. So networking is an essential element of this. And so we have to have the network. And by the way, the network now we should talk about is getting to be very sophisticated, not just on scale up and scale out, but also scale across. And so we just launched a capability for having multiple data centers hundreds of kilometers apart that can operate as one coherent computer. And so that's meaningful. The second area is on the security and safety side of the house, where there's a trust deficit. We're helping mitigate risks that both organizations and individuals might have when they use AI. And then the third one is a data gap. And we haven't talked about this as much, you know, in the past. But here's the thing with the data gap. 55% of the growth of data is machine data. Logs, metrics, events, traces. Now, imagine this. Let's say in the future you have 10 agents for every human. I'm being very conservative. I think there might be 100 agents for every human, right?

‍

Daniel Newman: Or more.

‍

Jeetu Patel: When you have that machine, data generated from those agents is going to be meaningfully higher than what it is right now. And so if you look at the curve of growth of data, your human data is going to grow at a certain curve. Your machine data is going to grow at a much steeper Curve. Right now, machine data doesn't correlate with human generated data publicly available on the Internet. To be able to go out and do better detections, better investigation, better figuring out when there's a breach that happens or not managing your infrastructure, all of that stuff requires machine data. And what Splunk has given us is the platform for machine data that every organization can use to make their data their moat, which everyone wants to do, but they don't really know how to do it. So that's what we can really help with. So we've got an open source, open weights, time series model. We've got a machine data lake that we've just announced. All of these things make a meaningful difference in how organizations are going to use data to harness the fullest potential of AI.

‍

Daniel Newman: So how's the. Just quickly, you know. Splunk was one of the biggest bets Cisco's ever made. Yeah, it sounds like that data corpus was a big part of what you got there. How's that going? Because I think Cisco's been at times underestimated in its security portfolio, but in my eyes, it's one of the best out there. And Splunk was kind of a missing piece. I mean, that part of your business has to really be seeing some pretty explosive growth.

‍

Jeetu Patel: It's actually one of the best. Splunk and Isovalent are two of the best acquisitions we've made. It's just so strategic to what we're doing. Of course, other acquisitions are great too, but Splunk and Isovalent are so stepping central to what we're doing right now in the AI era. And the reason for that is data is at the core, and the infrastructure to go out and manage data is great, but data is a very broad term. So what we've done is we've very intentionally focused on machine data. That's the superpower that Splunk has. We're not trying to go out and get to every piece of data everywhere. It's just machine data. What do we do to go out and drive efficient use of machine data, to correlate it with human generated data so that you can get better insights out of your data than what you could get otherwise? It's going really well.

‍

Daniel Newman: Yeah. So give me a bit of a macro viewpoint on Cisco and how you're thinking about enterprise data centers and compute. So we all know that, you know, you're pretty well the undisputed leader in networking. We just kind of hit on security briefly here. But you're more and more I'm seeing commitments to expand what you're doing on the compute side. Partnerships with Nvidia, expanded Edge compute offerings, expanded data center compute offerings. You'd been a little quiet for a period and it seems like you're back. It seems like Cisco's saying hey, we see this opportunity, we see that everything that can be built is being implemented. Kind of where's that going? And how do you not only get in there now while it's red hot, but how do you start, start to win, meaning like moving up the stack and being the first names talked about when it comes to building out the full data center compute.

‍

Jeetu Patel: So we are actually starting to see that happen now on an ongoing basis. So it's really gratifying to see where we were three years ago to now as a meaningful difference. It's a different company. There's a spring in the step for employees. The innovation has been at the fastest pace ever. Let's talk about the three problems that we're looking to solve at Cisco. Problem number one is to build out AI ready data centers for our customers. What does that mean? That means that we're going to provide compute, we're going to provide networking, we're going to provide security, we're going to provide optics. And it's not just for scale out networking where you have clusters of GPUs within a data center that need to connect. Because we make our own silicon, which you know really well, we are now able to go out and provide high performance solutions for data center interconnects where you can have a training run across multiple data centers that are hundreds of kilometers apart like I mentioned earlier. And so we just launched a P200 chip, a silicon kind of network ASIC and that P200 chip is 65% more power efficient and is a 51.2 terabit chip that when you put it into our system, which is the 8223 router, you can now connect multiple data centers that are hundreds of kilometers apart to act coherently. Why is that important? Because these models like years ago would sit on a single gpu. Then the model got big enough, they're like oh, I can't sit on a single GPU. I need to sit on a server with four to eight GPUs. Then it got bigger than that so we started having racks of server clusters. Once you had a rack that needed to be interconnected, that was a scale up networking. Now what you're starting to see is multiple racks need to get connected within a data center. And then what you're starting to find is a single data center can no longer host a massive training run because you might need multiple different data centers. Amazon just announced that they're going to have a million titanium GPUs for anthropic. Google just announced that they're actually launching, they're actually processing 1.3 quadrillion tokens a month. So the volumes are going up at an exponential curve where you're going to need to have all of these things get networked. And because we make our own silicon, we make our own systems, we make our own software, there's only a few companies that do this in the world. There's Broadcom, there's us, there's some GPU companies that do that. So there's only a few companies that do this in the world. So we're at the center of all of this movement right now.

‍

Daniel Newman: If I add a point to what you just said, part of the reason connecting data centers to data centers is getting gigawatt power to a single location is going to be complicated. It will happen in some places, but it's going to take a long time. So where you might have 100 megawatt days, or 200, you say, we can connect a 200 to a 200 to 100 to a 300, if you could have. And now we get almost a gigawatt across a physical, you know, space. I think that's an interesting way to make a gigawatt data center connecting four different ones that are in different locations.

‍

Jeetu Patel: And then the next big piece that'll happen is inferencing at the edge. And so when you start thinking about, in the enterprise especially, you start thinking about a branch office, you start thinking about a factory floor, you think about a sports center or a stadium, all of those places are going to need to have edge solutions. And so we just launched a Cisco unified Edge platform, which basically has networking, security, observability, all on the edge, where you don't actually have people. If you're in a hospital, you might not have local IT staff to manage that hospital as much. And so what we need to do is take out the complexity and make infrastructure at the edge plug and play. And so we just launched a solution where you can now take this one box which has everything connected in it, and go start running inferencing on the edge. And that's going to get even more important as you have more robotics, as you have remote surgeries, as you have, you know, kind of branch offices that need to do more inferencing on the edge for customer support, use cases. All of those things get to be really, really interesting.

‍

Daniel Newman: Yeah, well, we know that physical AI, by the way, is estimated to be a bigger opportunity than data center AI. It is, you know. And you're saying, where's the next trillion dollar opportunity? Well, there it is.

‍

Jeetu Patel: Several trillion, I would think, yes.

‍

Daniel Newman: I should say where's the next trillion plus dollar opportunity? It's very exciting. So let's wrap up here. I'd like to just get your read on how fast this goes for enterprise and sovereign. So we started with this kind of consumer versus enterprise and we started with kind of like the way AI is being read in the market is primarily based on ChatGPT usage. That's kind of the for dummies version of what's happening for those of us that are closer to it. We see every enterprise on the planet wants to engage with this stuff. That 95% MIT study, I actually think that was BS. I think, I think in theory, yes, every company's tried something and it hasn't worked. And then by the way, a Wharton study came out just after that said 75% of companies are having ROI weeks apart. It kind of goes back to Jeetu to just how you ask a question in many ways. But every company wants to get more efficient, they want to get more productive, they want to get more secure. They want to make sure that they've secured their business, not only secured cyber, but secured their business's future by taking advantage of the benefits I can give them. How fast does this start to proliferate? What drives enterprise to go from experimentation, proof of concept? Because I feel like we talked about this for a while to fully deploy solutions that are agents running their businesses and automations that are driving their workflows. Like how close do you think we are?

Jeetu Patel: I think it'll depend on the sector, segment and geography. Okay, so for example, pharmaceuticals are actually doing a pretty good job in using AI for a bunch of different use cases for drug development. You look at what's happening in healthcare, there's going to be a fair amount of progress that gets made. You look at what's happening in financial services, there's a fair amount of progress being made over there. And then there's also going to be two things that can be true at the same time. There might be a lot of experiments that fail and you might have a few experiments that people double down on that really, really succeed. And both those things can happen simultaneously, you know, and so it might not be inaccurate to say 90% of the experiments fail, but the 10% of the experiments that succeeded might actually have an exponential growth curve. And so you have to kind of look at it that way. And it's going to depend on the sector and it's also going to depend on the geography, you know, which part of the world is what happening in. Because I feel like over time what happens in these kinds of waves is you wait for a while, you wait for a while, you wait for a while, and then you see a big steep exponent that happens after. And that's exactly what's happening in the enterprise right now in a bunch of use cases. I'll give you a couple examples where this is very potent. Today, development in the enterprise with software engineers will almost 100% be augmented with AI in the next 24 months. It'll be unfathomable for you to think about an engineer that doesn't use AI. They will be an irrelevant engineer over time. Right. Customer support will be something that you will expect to make sure that there's an augmentation of human and an AI agent. And so these are things that are actually starting to gain a fair amount of momentum. And I think the thing that's going to hold us back is actually the cost of infrastructure, is the availability of infrastructure there. Because enterprises right now will not have the power to go out and build out their own data centers. They're going to go to hyperscalers to do that initially, and then over time they'll start building out their own. And so that's where we have to actually start thinking about whether the constraints will dictate the conditions of the market far more than the demand signal.

‍

Daniel Newman: But if that is true, then effectively what that means is that there really is no bubble for AI, because if that demand is there, meaning that if there are enough enterprises that want more access, that they need more compute. I'm not saying there aren't outliers, I'm not saying there aren't companies that get dragged along in a sort of exuberant market. But what I'm saying is when people say, is this all going to get used up? These are not ghost cities. These data centers that are being built, if I'm hearing correct, are going to.

‍

Jeetu Patel: Be my opinion, and this is just an opinion, but in my opinion, we are underestimating the demand signal for infrastructure and AI. And in my opinion, there's also going to be some companies that have frothy valuations. But is this the largest movement we will have seen the largest platform shift we have seen in our lifetimes by a long shot. This is way bigger than mobile. This is way bigger than the Internet. This is way bigger than any other revolution we've seen in the cloud, any of those. And it's actually a step function higher than any of those. But that doesn't mean that you won't have frothy valuations from companies.

‍

Daniel Newman: Oh, of course.

‍

Jeetu Patel: And I think what people confuse sometimes is frothy valuations is a precondition of a hyper growth industry because a lot of experiments get tried out and people.

‍

Daniel Newman: Want to race in. They don't.

‍

Jeetu Patel: But there will be new industries created, there will be massive levels of investments that will be made and you will see huge amounts of growth economically as a result of AI that will contribute to society.

‍

Daniel Newman: The answer is, and I'll leave it here, is that the investors would rather be early than late. And what I'm saying is they might be rather overpaid now and knowing these companies might grow into those numbers later than bet on waiting until they actually get to those numbers. And it's a little bit like what you said about the infrastructure. They'd rather build too much and let the demand come later.

‍

Jeetu Patel: The risk is just too high to sit on the sidelines right now. And I think, I think here's what will happen. A lot of companies that sit on the sidelines will become irrelevant.

‍

Daniel Newman: Well, it's good to see that you are driving that kind of change at Cisco and you're not sitting on the sidelines. It's been a great year too. You know, you and I have talked a lot about it and you know, you guys started kind of, you observed, observed and then you've gone full tilt and you've gone really fast. And I saw a lot of great innovations.

Jeetu Patel: We're all in. We're the pig and the breakfast man.

‍

Daniel Newman: And I appreciate you even leaving me a little bit of sausage, a little bit of bacon to eat here. But G2, let's, let's do this again soon. Great being with you here, buddy.

Jeetu Patel: Thank you, man.

‍

Daniel Newman: Thanks for joining the show.

‍

Jeetu Patel: Thanks for having me.

‍

Daniel Newman: Thank you everybody for joining us for this Six Five On The Road. Great conversation there. Regular here on the Six Five Jeetu Patel hit subscribe. Join us for all of our other content here on the. On the platform. We appreciate you tuning in. We'll see you all later.

‍